Multi-Class Image Classification Model for Belgium Traffic Sign Using TensorFlow Take 2¶

David Lowe¶

June 21, 2022¶

SUMMARY: This project aims to construct a predictive model using a TensorFlow convolutional neural network (CNN) and document the end-to-end steps using a template. The Belgium Traffic Sign dataset is a multi-class classification situation where we attempt to predict one of several (more than two) possible outcomes.

INTRODUCTION: This dataset contains over 7,200 images of 62 varieties of traffic signs used in Belgium. The researcher performed experiments on the dataset to create a CNN-based classification system.

ANALYSIS: The InceptionV3 model's performance achieved an accuracy score of 98.84% after 10 epochs using the training dataset. When we applied the model to the validation dataset, the model achieved an accuracy score of 93.73%.

CONCLUSION: In this iteration, the TensorFlow InceptionV3 CNN model appeared suitable for modeling this dataset.

Dataset ML Model: Multi-Class classification with numerical features

Dataset Used: Belgium_Traffic_Sign_image_data_62_class_data

Dataset Reference: https://www.kaggle.com/datasets/abhi8923shriv/belgium-ts

One source of potential performance benchmarks: https://www.kaggle.com/datasets/abhi8923shriv/belgium-ts/code

Task 1 - Prepare Environment¶

In [1]:
# Retrieve CPU information from the system
ncpu = !nproc
print("The number of available CPUs is:", ncpu[0])

The number of available CPUs is: 2
In [2]:
# Retrieve memory configuration information
from psutil import virtual_memory
ram_gb = virtual_memory().total / 1e9
print('Your runtime has {:.1f} gigabytes of available RAM\n'.format(ram_gb))

Your runtime has 13.6 gigabytes of available RAM
In [3]:
# Retrieve GPU configuration information
gpu_info = !nvidia-smi
gpu_info = '\n'.join(gpu_info)
print(gpu_info)

Mon Jun 13 01:13:17 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.32.03    Driver Version: 460.32.03    CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla V100-SXM2...  Off  | 00000000:00:04.0 Off |                    0 |
| N/A   33C    P0    22W / 300W |      0MiB / 16160MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

1.a) Load libraries and modules¶

In [4]:
# Set the random seed number for reproducible results
RNG_SEED = 888
In [5]:
import random
random.seed(RNG_SEED)
import numpy as np
np.random.seed(RNG_SEED)
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import os
import sys
import math
# import boto3
import zipfile
from datetime import datetime
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score

import tensorflow as tf
tf.random.set_seed(RNG_SEED)
from tensorflow import keras
from tensorflow.keras.callbacks import ReduceLROnPlateau
from tensorflow.keras.preprocessing.image import ImageDataGenerator

1.b) Set up the controlling parameters and functions¶

In [6]:
# Begin the timer for the script processing
START_TIME_SCRIPT = datetime.now()
In [7]:
# Set up the number of CPU cores available for multi-thread processing
N_JOBS = 1

# Set up the flag to stop sending progress emails (setting to True will send status emails!)
NOTIFY_STATUS = False

# Set the percentage sizes for splitting the dataset
TEST_SET_RATIO = 0.1
VAL_SET_RATIO = 0.1

# Set the number of folds for cross validation
N_FOLDS = 5
N_ITERATIONS = 1

# Set various default modeling parameters
DEFAULT_LOSS = 'categorical_crossentropy'
DEFAULT_METRICS = ['accuracy']
DEFAULT_OPTIMIZER = tf.keras.optimizers.Adam(learning_rate=0.00001)
CLASSIFIER_ACTIVATION = 'softmax'
MAX_EPOCHS = 10
BATCH_SIZE = 32
NUM_CLASSES = 62
# CLASS_LABELS = []
# CLASS_NAMES = []
# RAW_IMAGE_SIZE = (250, 250)
TARGET_IMAGE_SIZE = (299, 299)
INPUT_IMAGE_SHAPE = (TARGET_IMAGE_SIZE[0], TARGET_IMAGE_SIZE[1], 3)

# Define the labels to use for graphing the data
TRAIN_METRIC = "accuracy"
VALIDATION_METRIC = "val_accuracy"
TRAIN_LOSS = "loss"
VALIDATION_LOSS = "val_loss"

# Define the directory locations and file names
STAGING_DIR = 'staging/'
TRAIN_DIR = 'staging/BelgiumTSC_Training/Training/'
VALID_DIR = 'staging/BelgiumTSC_Testing/Testing/'
TEST_DIR = ''
TRAIN_DATASET = 'belgium-traffic-sign.zip'
# VALID_DATASET = ''
# TEST_DATASET = ''
# TRAIN_LABELS = ''
# VALID_LABELS = ''
# TEST_LABELS = ''
# OUTPUT_DIR = 'staging/'
# SAMPLE_SUBMISSION_CSV = 'sample_submission.csv'
# FINAL_SUBMISSION_CSV = 'submission.csv'

# Check the number of GPUs accessible through TensorFlow
print('Num GPUs Available:', len(tf.config.list_physical_devices('GPU')))

# Print out the TensorFlow version for confirmation
print('TensorFlow version:', tf.__version__)

Num GPUs Available: 1
TensorFlow version: 2.8.2
In [8]:
# Set up the email notification function
def status_notify(msg_text):
    access_key = os.environ.get('SNS_ACCESS_KEY')
    secret_key = os.environ.get('SNS_SECRET_KEY')
    aws_region = os.environ.get('SNS_AWS_REGION')
    topic_arn = os.environ.get('SNS_TOPIC_ARN')
    if (access_key is None) or (secret_key is None) or (aws_region is None):
        sys.exit("Incomplete notification setup info. Script Processing Aborted!!!")
    sns = boto3.client('sns', aws_access_key_id=access_key, aws_secret_access_key=secret_key, region_name=aws_region)
    response = sns.publish(TopicArn=topic_arn, Message=msg_text)
    if response['ResponseMetadata']['HTTPStatusCode'] != 200 :
        print('Status notification not OK with HTTP status code:', response['ResponseMetadata']['HTTPStatusCode'])
In [9]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 1 - Prepare Environment completed on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))

Task 2 - Load and Prepare Images¶

In [10]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 2 - Load and Prepare Images has begun on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))
In [11]:
# Clean up the old files and download directories before receiving new ones
!rm -rf staging/
# !rm archive.zip
!mkdir staging/
In [12]:
if not os.path.exists(TRAIN_DATASET):
    !wget https://dainesanalytics.com/datasets/kaggle-abhi8923shriv-belgium-traffic-sign/belgium-traffic-sign.zip

--2022-06-13 01:13:24--  https://dainesanalytics.com/datasets/kaggle-abhi8923shriv-belgium-traffic-sign/belgium-traffic-sign.zip
Resolving dainesanalytics.com (dainesanalytics.com)... 65.9.85.3, 65.9.85.28, 65.9.85.71, ...
Connecting to dainesanalytics.com (dainesanalytics.com)|65.9.85.3|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 452884286 (432M) [application/zip]
Saving to: ‘belgium-traffic-sign.zip’

belgium-traffic-sig 100%[===================>] 431.90M  26.2MB/s    in 20s     

2022-06-13 01:13:45 (21.6 MB/s) - ‘belgium-traffic-sign.zip’ saved [452884286/452884286]
In [13]:
zip_ref = zipfile.ZipFile(TRAIN_DATASET, 'r')
zip_ref.extractall(STAGING_DIR)
zip_ref.close()
In [14]:
# Delete all the annotation files embedded in the image folders
!rm staging/BelgiumTSC_Training/Training/*.txt
!find staging/BelgiumTSC_Training/ -name "*csv" | xargs rm
!rm staging/BelgiumTSC_Testing/Testing/*.txt
!find staging/BelgiumTSC_Testing/ -name "*csv" | xargs rm
In [15]:
CLASS_LABELS = os.listdir(TRAIN_DIR)
print(CLASS_LABELS)

['00017', '00025', '00002', '00056', '00014', '00013', '00038', '00004', '00020', '00031', '00045', '00028', '00057', '00015', '00046', '00024', '00018', '00034', '00009', '00043', '00005', '00019', '00061', '00049', '00012', '00010', '00059', '00052', '00058', '00029', '00055', '00023', '00053', '00027', '00022', '00011', '00021', '00048', '00033', '00037', '00003', '00040', '00030', '00047', '00026', '00039', '00036', '00044', '00050', '00054', '00060', '00042', '00006', '00051', '00000', '00016', '00035', '00008', '00007', '00001', '00032', '00041']
In [16]:
# Brief listing of training image files for each class
for c_label in CLASS_LABELS:
    training_class_dir = os.path.join(TRAIN_DIR, c_label)
    training_class_files = os.listdir(training_class_dir)
    print('Number of training images for', c_label, ':', len(os.listdir(training_class_dir)))
    print('Training samples for', c_label, ':', training_class_files[:5],'\n')

Number of training images for 00017 : 79
Training samples for 00017 : ['00253_00001.ppm', '00845_00001.ppm', '01199_00001.ppm', '00910_00002.ppm', '00156_00001.ppm'] 

Number of training images for 00025 : 42
Training samples for 00025 : ['01151_00002.ppm', '01166_00002.ppm', '00696_00001.ppm', '00701_00001.ppm', '01184_00000.ppm'] 

Number of training images for 00002 : 13
Training samples for 00002 : ['01722_00002.ppm', '01503_00001.ppm', '01320_00000.ppm', '01722_00000.ppm', '01515_00001.ppm'] 

Number of training images for 00056 : 95
Training samples for 00056 : ['00296_00002.ppm', '01199_00001.ppm', '00784_00001.ppm', '00256_00001.ppm', '00784_00000.ppm'] 

Number of training images for 00014 : 43
Training samples for 00014 : ['00448_00001.ppm', '00408_00001.ppm', '00899_00000.ppm', '01925_00000.ppm', '01544_00000.ppm'] 

Number of training images for 00013 : 90
Training samples for 00013 : ['00866_00002.ppm', '01704_00002.ppm', '00869_00001.ppm', '00235_00000.ppm', '00884_00000.ppm'] 

Number of training images for 00038 : 285
Training samples for 00038 : ['00004_00002.ppm', '00796_00001.ppm', '00192_00002.ppm', '00686_00000.ppm', '01382_00001.ppm'] 

Number of training images for 00004 : 15
Training samples for 00004 : ['00145_00001.ppm', '00623_00000.ppm', '00214_00002.ppm', '01312_00001.ppm', '00623_00001.ppm'] 

Number of training images for 00020 : 42
Training samples for 00020 : ['01968_00002.ppm', '01968_00000.ppm', '01755_00002.ppm', '00332_00002.ppm', '01859_00002.ppm'] 

Number of training images for 00031 : 63
Training samples for 00031 : ['01909_00002.ppm', '01957_00002.ppm', '00272_00000.ppm', '01957_00001.ppm', '01716_00002.ppm'] 

Number of training images for 00045 : 74
Training samples for 00045 : ['00035_00001.ppm', '00833_00001.ppm', '00451_00001.ppm', '01637_00000.ppm', '01392_00002.ppm'] 

Number of training images for 00028 : 125
Training samples for 00028 : ['01896_00002.ppm', '00662_00001.ppm', '00155_00002.ppm', '01897_00001.ppm', '00040_00001.ppm'] 

Number of training images for 00057 : 78
Training samples for 00057 : ['01271_00002.ppm', '00472_00002.ppm', '00784_00001.ppm', '00784_00000.ppm', '00339_00000.ppm'] 

Number of training images for 00015 : 9
Training samples for 00015 : ['00675_00001.ppm', '00675_00000.ppm', '00709_00001.ppm', '00709_00000.ppm', '00675_00002.ppm'] 

Number of training images for 00046 : 44
Training samples for 00046 : ['01021_00002.ppm', '01628_00002.ppm', '00090_00001.ppm', '01116_00002.ppm', '00814_00000.ppm'] 

Number of training images for 00024 : 48
Training samples for 00024 : ['01845_00001.ppm', '01952_00000.ppm', '00355_00001.ppm', '01870_00002.ppm', '01280_00001.ppm'] 

Number of training images for 00018 : 81
Training samples for 00018 : ['00008_00000.ppm', '00365_00001.ppm', '00009_00000.ppm', '00429_00000.ppm', '01087_00002.ppm'] 

Number of training images for 00034 : 46
Training samples for 00034 : ['01418_00002.ppm', '01240_00001.ppm', '01260_00002.ppm', '01241_00001.ppm', '01646_00001.ppm'] 

Number of training images for 00009 : 18
Training samples for 00009 : ['00410_00002.ppm', '00141_00000.ppm', '00285_00001.ppm', '00285_00002.ppm', '00141_00001.ppm'] 

Number of training images for 00043 : 30
Training samples for 00043 : ['01589_00001.ppm', '00073_00000.ppm', '01735_00001.ppm', '00873_00001.ppm', '00878_00002.ppm'] 

Number of training images for 00005 : 11
Training samples for 00005 : ['00261_00000.ppm', '00261_00001.ppm', '00575_00000.ppm', '00390_00001.ppm', '00575_00002.ppm'] 

Number of training images for 00019 : 231
Training samples for 00019 : ['00923_00000.ppm', '00126_00002.ppm', '01344_00001.ppm', '00983_00002.ppm', '01355_00002.ppm'] 

Number of training images for 00061 : 282
Training samples for 00061 : ['01909_00002.ppm', '00192_00002.ppm', '01957_00002.ppm', '00951_00002.ppm', '01269_00001.ppm'] 

Number of training images for 00049 : 12
Training samples for 00049 : ['00311_00000.ppm', '00310_00000.ppm', '00720_00001.ppm', '00720_00000.ppm', '00258_00002.ppm'] 

Number of training images for 00012 : 18
Training samples for 00012 : ['01565_00002.ppm', '01565_00001.ppm', '01473_00002.ppm', '01473_00000.ppm', '01609_00001.ppm'] 

Number of training images for 00010 : 21
Training samples for 00010 : ['01903_00000.ppm', '01903_00001.ppm', '00289_00002.ppm', '01903_00002.ppm', '00263_00002.ppm'] 

Number of training images for 00059 : 42
Training samples for 00059 : ['00948_00002.ppm', '01586_00001.ppm', '01481_00001.ppm', '00350_00001.ppm', '01583_00000.ppm'] 

Number of training images for 00052 : 27
Training samples for 00052 : ['01131_00001.ppm', '00476_00002.ppm', '01128_00001.ppm', '00901_00002.ppm', '00476_00001.ppm'] 

Number of training images for 00058 : 15
Training samples for 00058 : ['01444_00000.ppm', '01444_00001.ppm', '00107_00001.ppm', '00776_00001.ppm', '00107_00000.ppm'] 

Number of training images for 00029 : 33
Training samples for 00029 : ['00376_00002.ppm', '01541_00002.ppm', '00413_00001.ppm', '00909_00002.ppm', '00021_00001.ppm'] 

Number of training images for 00055 : 12
Training samples for 00055 : ['00191_00001.ppm', '00276_00000.ppm', '01905_00001.ppm', '00191_00002.ppm', '01905_00002.ppm'] 

Number of training images for 00023 : 15
Training samples for 00023 : ['00535_00001.ppm', '00465_00001.ppm', '00535_00002.ppm', '00630_00000.ppm', '00785_00002.ppm'] 

Number of training images for 00053 : 199
Training samples for 00053 : ['01658_00001.ppm', '01554_00000.ppm', '01554_00001.ppm', '01052_00000.ppm', '01793_00002.ppm'] 

Number of training images for 00027 : 18
Training samples for 00027 : ['01597_00000.ppm', '01544_00000.ppm', '01544_00002.ppm', '01667_00002.ppm', '01597_00002.ppm'] 

Number of training images for 00022 : 375
Training samples for 00022 : ['01655_00002.ppm', '00840_00001.ppm', '00936_00001.ppm', '01055_00002.ppm', '01155_00000.ppm'] 

Number of training images for 00011 : 7
Training samples for 00011 : ['01507_00002.ppm', '00271_00003.ppm', '00271_00000.ppm', '01507_00001.ppm', '00271_00001.ppm'] 

Number of training images for 00021 : 43
Training samples for 00021 : ['01832_00000.ppm', '00684_00000.ppm', '01868_00001.ppm', '01258_00001.ppm', '00554_00002.ppm'] 

Number of training images for 00048 : 11
Training samples for 00048 : ['00128_00000.ppm', '00128_00001.ppm', '00091_00000.ppm', '00091_00002.ppm', '01923_00001.ppm'] 

Number of training images for 00033 : 12
Training samples for 00033 : ['00339_00000.ppm', '00343_00002.ppm', '00339_00002.ppm', '00343_00001.ppm', '00340_00001.ppm'] 

Number of training images for 00037 : 98
Training samples for 00037 : ['01344_00001.ppm', '01355_00002.ppm', '00975_00002.ppm', '00163_00001.ppm', '01346_00002.ppm'] 

Number of training images for 00003 : 15
Training samples for 00003 : ['00211_00000.ppm', '00211_00001.ppm', '00391_00000.ppm', '01291_00000.ppm', '00207_00001.ppm'] 

Number of training images for 00040 : 242
Training samples for 00040 : ['01791_00002.ppm', '01177_00001.ppm', '01411_00000.ppm', '00448_00001.ppm', '00829_00000.ppm'] 

Number of training images for 00030 : 37
Training samples for 00030 : ['00686_00000.ppm', '00405_00001.ppm', '01293_00000.ppm', '00625_00000.ppm', '00328_00000.ppm'] 

Number of training images for 00047 : 147
Training samples for 00047 : ['00487_00000.ppm', '01410_00002.ppm', '01596_00000.ppm', '00498_00001.ppm', '01694_00002.ppm'] 

Number of training images for 00026 : 6
Training samples for 00026 : ['00246_00002.ppm', '00173_00000.ppm', '00246_00000.ppm', '00173_00001.ppm', '00173_00002.ppm'] 

Number of training images for 00039 : 196
Training samples for 00039 : ['00923_00000.ppm', '01909_00002.ppm', '01838_00001.ppm', '00944_00001.ppm', '00626_00001.ppm'] 

Number of training images for 00036 : 18
Training samples for 00036 : ['01687_00000.ppm', '00564_00000.ppm', '01285_00000.ppm', '00564_00002.ppm', '01457_00001.ppm'] 

Number of training images for 00044 : 48
Training samples for 00044 : ['01966_00000.ppm', '01597_00000.ppm', '00228_00002.ppm', '00228_00001.ppm', '00289_00002.ppm'] 

Number of training images for 00050 : 15
Training samples for 00050 : ['00320_00001.ppm', '00824_00000.ppm', '00581_00001.ppm', '00322_00000.ppm', '00584_00002.ppm'] 

Number of training images for 00054 : 118
Training samples for 00054 : ['01757_00001.ppm', '01415_00002.ppm', '00489_00001.ppm', '01937_00001.ppm', '00162_00002.ppm'] 

Number of training images for 00060 : 9
Training samples for 00060 : ['01892_00000.ppm', '01877_00002.ppm', '00209_00002.ppm', '01892_00001.ppm', '00209_00000.ppm'] 

Number of training images for 00042 : 35
Training samples for 00042 : ['00076_00000.ppm', '00076_00002.ppm', '01582_00002.ppm', '01582_00000.ppm', '01166_00002.ppm'] 

Number of training images for 00006 : 18
Training samples for 00006 : ['00220_00000.ppm', '00215_00001.ppm', '00215_00000.ppm', '00326_00001.ppm', '00327_00002.ppm'] 

Number of training images for 00051 : 27
Training samples for 00051 : ['01127_00002.ppm', '01878_00000.ppm', '01761_00000.ppm', '00902_00002.ppm', '00470_00000.ppm'] 

Number of training images for 00000 : 15
Training samples for 00000 : ['01797_00000.ppm', '01797_00002.ppm', '01798_00002.ppm', '01797_00001.ppm', '01799_00002.ppm'] 

Number of training images for 00016 : 9
Training samples for 00016 : ['00444_00001.ppm', '00332_00002.ppm', '00444_00002.ppm', '00444_00000.ppm', '00332_00001.ppm'] 

Number of training images for 00035 : 60
Training samples for 00035 : ['00378_00002.ppm', '00345_00001.ppm', '00588_00000.ppm', '01711_00001.ppm', '00345_00002.ppm'] 

Number of training images for 00008 : 27
Training samples for 00008 : ['01578_00001.ppm', '01016_00002.ppm', '00046_00001.ppm', '01520_00001.ppm', '01578_00002.ppm'] 

Number of training images for 00007 : 157
Training samples for 00007 : ['01801_00000.ppm', '01433_00000.ppm', '01159_00002.ppm', '00748_00001.ppm', '00618_00002.ppm'] 

Number of training images for 00001 : 110
Training samples for 00001 : ['00801_00002.ppm', '01103_00002.ppm', '00468_00002.ppm', '01651_00000.ppm', '00806_00000.ppm'] 

Number of training images for 00032 : 316
Training samples for 00032 : ['01848_00000.ppm', '00192_00002.ppm', '01818_00002.ppm', '01849_00000.ppm', '01828_00000.ppm'] 

Number of training images for 00041 : 148
Training samples for 00041 : ['01778_00000.ppm', '01078_00000.ppm', '01371_00002.ppm', '01449_00000.ppm', '00889_00001.ppm']
In [17]:
# Plot some training images from the dataset
nrows = len(CLASS_LABELS)
ncols = 4
training_examples = []
example_labels = []

fig = plt.gcf()
fig.set_size_inches(ncols * 4, nrows * 3)

for c_label in CLASS_LABELS:
    training_class_dir = os.path.join(TRAIN_DIR, c_label)
    training_class_files = os.listdir(training_class_dir)
    for j in range(ncols):
        training_examples.append(training_class_dir + '/' + training_class_files[j])
        example_labels.append(c_label)
    # print(training_examples)
    # print(example_labels)

for i, img_path in enumerate(training_examples):
    # Set up subplot; subplot indices start at 1
    sp = plt.subplot(nrows, ncols, i+1)
    sp.text(0, 0, example_labels[i])
    # sp.axis('Off')
    img = mpimg.imread(img_path)
    plt.imshow(img)
plt.show()
In [18]:
datagen_kwargs = dict(rescale=1./255)
training_datagen = ImageDataGenerator(**datagen_kwargs)
validation_datagen = ImageDataGenerator(**datagen_kwargs)
dataflow_kwargs = dict(class_mode="categorical")

do_data_augmentation = True
if do_data_augmentation:
    training_datagen = ImageDataGenerator(rotation_range=45,
                                          horizontal_flip=True,
                                          vertical_flip=True,
                                          **datagen_kwargs)

print('Loading and pre-processing the training images...')
training_generator = training_datagen.flow_from_directory(directory=TRAIN_DIR,
                                                          target_size=TARGET_IMAGE_SIZE,
                                                          batch_size=BATCH_SIZE,
                                                          shuffle=True,
                                                          seed=RNG_SEED,
                                                          **dataflow_kwargs)
print('Number of training image batches per epoch of modeling:', len(training_generator))

print('Loading and pre-processing the validation images...')
validation_generator = validation_datagen.flow_from_directory(directory=VALID_DIR,
                                                              target_size=TARGET_IMAGE_SIZE,
                                                              batch_size=BATCH_SIZE,
                                                              shuffle=False,
                                                              **dataflow_kwargs)
print('Number of validation image batches per epoch of modeling:', len(validation_generator))

Loading and pre-processing the training images...
Found 4575 images belonging to 62 classes.
Number of training image batches per epoch of modeling: 143
Loading and pre-processing the validation images...
Found 2520 images belonging to 62 classes.
Number of validation image batches per epoch of modeling: 79
In [19]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 2 - Load and Prepare Images completed on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))

Task 3 - Define and Train Models¶

In [20]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 3 - Define and Train Models has begun on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))
In [21]:
# Define the function for plotting training results for comparison
def plot_metrics(history):
    fig, axs = plt.subplots(1, 2, figsize=(24, 15))
    metrics =  [TRAIN_LOSS, TRAIN_METRIC]
    for n, metric in enumerate(metrics):
        name = metric.replace("_"," ").capitalize()
        plt.subplot(2,2,n+1)
        plt.plot(history.epoch, history.history[metric], color='blue', label='Train')
        plt.plot(history.epoch, history.history['val_'+metric], color='red', linestyle="--", label='Val')
        plt.xlabel('Epoch')
        plt.ylabel(name)
        if metric == TRAIN_LOSS:
            plt.ylim([0, plt.ylim()[1]])
        else:
            plt.ylim([0, 1])
        plt.legend()
In [22]:
# Define the baseline model for benchmarking
def create_nn_model(input_param=INPUT_IMAGE_SHAPE, output_param=NUM_CLASSES, dense_nodes=2048,
                    classifier_activation=CLASSIFIER_ACTIVATION, loss_param=DEFAULT_LOSS,
                    opt_param=DEFAULT_OPTIMIZER, metrics_param=DEFAULT_METRICS):
    base_model = keras.applications.inception_v3.InceptionV3(include_top=False, weights='imagenet', input_shape=input_param)
    nn_model = keras.models.Sequential()
    nn_model.add(base_model)
    nn_model.add(keras.layers.Flatten())
    nn_model.add(keras.layers.Dense(dense_nodes, activation='relu')),
    nn_model.add(keras.layers.Dense(output_param, activation=classifier_activation))
    nn_model.compile(loss=loss_param, optimizer=opt_param, metrics=metrics_param)
    return nn_model
In [23]:
# Initialize the neural network model and get the training results for plotting graph
start_time_module = datetime.now()
tf.keras.utils.set_random_seed(RNG_SEED)
baseline_model = create_nn_model()
baseline_model_history = baseline_model.fit(training_generator,
                                            epochs=MAX_EPOCHS,
                                            validation_data=validation_generator,
                                            verbose=1)
print('Total time for model fitting:', (datetime.now() - start_time_module))

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/inception_v3/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5
87916544/87910968 [==============================] - 1s 0us/step
87924736/87910968 [==============================] - 1s 0us/step
Epoch 1/10
143/143 [==============================] - 118s 694ms/step - loss: 1.7377 - accuracy: 0.5952 - val_loss: 1.0416 - val_accuracy: 0.7325
Epoch 2/10
143/143 [==============================] - 95s 661ms/step - loss: 0.5367 - accuracy: 0.8533 - val_loss: 0.6151 - val_accuracy: 0.8317
Epoch 3/10
143/143 [==============================] - 96s 671ms/step - loss: 0.2811 - accuracy: 0.9169 - val_loss: 0.5905 - val_accuracy: 0.8333
Epoch 4/10
143/143 [==============================] - 96s 669ms/step - loss: 0.1772 - accuracy: 0.9475 - val_loss: 0.3786 - val_accuracy: 0.8976
Epoch 5/10
143/143 [==============================] - 95s 667ms/step - loss: 0.1377 - accuracy: 0.9607 - val_loss: 0.2826 - val_accuracy: 0.9266
Epoch 6/10
143/143 [==============================] - 96s 673ms/step - loss: 0.1109 - accuracy: 0.9679 - val_loss: 0.2974 - val_accuracy: 0.9250
Epoch 7/10
143/143 [==============================] - 96s 669ms/step - loss: 0.0774 - accuracy: 0.9770 - val_loss: 0.2436 - val_accuracy: 0.9345
Epoch 8/10
143/143 [==============================] - 100s 699ms/step - loss: 0.0816 - accuracy: 0.9766 - val_loss: 0.2530 - val_accuracy: 0.9310
Epoch 9/10
143/143 [==============================] - 97s 675ms/step - loss: 0.0693 - accuracy: 0.9808 - val_loss: 0.3889 - val_accuracy: 0.9004
Epoch 10/10
143/143 [==============================] - 96s 669ms/step - loss: 0.0447 - accuracy: 0.9884 - val_loss: 0.2479 - val_accuracy: 0.9373
Total time for model fitting: 0:16:32.726621
In [24]:
baseline_model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 inception_v3 (Functional)   (None, 8, 8, 2048)        21802784  
                                                                 
 flatten (Flatten)           (None, 131072)            0         
                                                                 
 dense (Dense)               (None, 2048)              268437504 
                                                                 
 dense_1 (Dense)             (None, 62)                127038    
                                                                 
=================================================================
Total params: 290,367,326
Trainable params: 290,332,894
Non-trainable params: 34,432
_________________________________________________________________
In [25]:
plot_metrics(baseline_model_history)
In [26]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 3 - Define and Train Models completed on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))

Task 4 - Tune and Optimize Models¶

In [27]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 4 - Tune and Optimize Models has begun on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))
In [28]:
# Not applicable for this iteration of modeling
In [29]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 4 - Tune and Optimize Models completed on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))

Task 5 - Finalize Model and Make Predictions¶

In [30]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 5 - Finalize Model and Make Predictions has begun on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))
In [31]:
# Not applicable for this iteration of modeling
In [32]:
if NOTIFY_STATUS: status_notify('(TensorFlow Multi-Class) Task 5 - Finalize Model and Make Predictions completed on ' + datetime.now().strftime('%A %B %d, %Y %I:%M:%S %p'))
In [33]:
print ('Total time for the script:',(datetime.now() - START_TIME_SCRIPT))

Total time for the script: 0:17:28.556112